How to Implement AI Roleplay for Managers: HR Guide
A 12-week plan for HR teams rolling out AI roleplay to people leaders: privacy, first cohort, success measures, launch and review, with free templates.
Home › Resources › Implementing AI roleplay for managers
The short answer
To implement AI roleplay for managers: name the capability gap and a sponsor who'll go first, decide in writing what HR will and won't see, pick two or three conversations managers will have this quarter, start with a first cohort of roughly 20–60 people leaders, write your success measures before launch, time the launch to a real moment like a performance cycle, then review at weeks 4, 8 and 12 and decide whether to expand, adjust or stop.
Why most rollouts stall
The day a new tool launches is often the day it becomes everyone's optional extra. There's an all-staff email, a login, a burst of curiosity in week one, and by week six the only people still using it are the ones who didn't need it.
I don't think that's a technology problem. In my clinical work, people rarely change because they were given access to something. They change when the support arrives at the moment they need it, from someone they trust, without the sense of being watched. Rollouts of manager tools fail for the same reasons: they're launched at the wrong moment, nobody senior goes first, and managers aren't sure who's looking.
This guide is a plan for avoiding those three mistakes. It's written for Australian HR and People & Culture teams, but most of it travels.
The People leaders we speak with are consistent on one point: managers won't practise every day. They'll practise once, shortly before the conversation that matters. Design for that moment, not for a weekly habit target.
- Before you startPrivacy, cohort, measures
- Weeks 1–2Launch on a real moment
- Weeks 3–8Support the people, not a dashboard
- Weeks 9–12Review against the measures
- DecideExpand, adjust or stop
Shape of the rollout. Each block links to the section that holds the work.
Before you start: name the problem and the sponsor
Direct answer: write down the capability gap in one sentence, and find a senior leader who'll practise first and say so.
Name the gap. "Improve leadership capability" is too broad to measure. Try something closer to: "We can't tell whether our managers can handle the performance conversations that matter, and we only find out when something goes wrong." That sentence will shape every decision after it.
Find a sponsor who'll go first. The single biggest signal to managers is a senior leader saying, "I practised this before my own conversation, and it was harder than I expected." Top-down beats bottom-up for this kind of tool.
Build the business case in your numbers, not a vendor's.
Business case worksheet
| Question | Your answer |
|---|---|
| How many people leaders do we have? | |
| Which conversations do they most often avoid or get wrong? | |
| How many manager-related HR escalations did we handle last year? | |
| How many exits in the last 12 months mentioned the manager? | |
| Roughly how many HR hours go into coaching managers through conversations each month? | |
| What does leadership development currently cost per manager, and what evidence do we get back? | |
| What would we need to see at week 12 to keep going? |
Template. Select the table and copy it. No email.
Avoid leading the business case with "this saves HR time". Executives often hear that as HR making its own job easier. Lead with manager capability and the cost of conversations that go wrong.
Phase 1: Decide what HR will and won't see
Direct answer: agree in writing, before anyone logs in, what's private to the manager and what HR can see. Practice should never become an HR file.
This is the decision that makes or breaks adoption. If managers suspect HR is reading their practice, they'll either avoid the tool or play it safe, and play-it-safe practice is useless.
A trust-preserving default
| Private to the manager | Visible to HR |
|---|---|
| What they said in a practice conversation | Which conversation types are being practised |
| Their individual attempts and retries | Whether practice is reaching the agreed standard, at a group level |
| Their prep notes | Where the common gaps are (for example, "next steps rarely include a date") |
A default to agree before launch. It is not a description of any one product's admin screen.
Things to settle in this phase:
- Who in HR can see what, and at what level of detail.
- Whether any manager-level view exists. If it does, frame it as who might need extra support, never as a pass/fail list.
- How the tool handles data. Involve your privacy officer and IT early, and check the vendor's answers against your obligations under Australian privacy law.
- How you'll explain all this to managers in two sentences.
Write the privacy promise in plain language and put it in the launch message. For example: "What you say in practice stays with you. HR sees whether the organisation is getting better at these conversations, not what you said."
Phase 2: Choose the conversations and the first cohort
Direct answer: pick two or three conversations managers will have this quarter, and start with a defined group of roughly 20–60 people leaders.
Choose two or three conversations
Start with what's already in managers' calendars, not a catalogue.
Conversations to start with
| Conversation | Why start here | Timing tip |
|---|---|---|
| Performance feedback | The most common conversation, and the one most often avoided or blurred | Any time; especially before mid-year or end-of-year reviews |
| Coaching a high performer | Balances the program so it isn't seen as a performance-management tool | Pair with performance feedback from day one |
| Pay or promotion "no" | High emotion, predictable timing | Launch a few weeks before your remuneration review |
| Team conflict | Unresolved friction is costly and tends to escalate | When a team is going through change |
Don't launch with termination or formal performance-improvement conversations as the headline. They matter, but leading with them tells managers the tool is about managing people out, and adoption drops.
Choose the first cohort
- Size: roughly 20–60 people leaders. That's big enough to see patterns and small enough to support personally.
- Mix: new managers, experienced managers and at least one senior leader who practises visibly.
- Pull, not push: include managers with a known conversation coming up. They have a reason to use it this week.
- Avoid: making the first cohort a remedial group. If it's seen as being for "struggling managers", nobody else will touch it.
Choose a tool
Before signing, run the vendor through the eight questions in how to evaluate AI roleplay for manager training. The two that matter most at implementation are whether each conversation has its own standard, and who sees what.
Phase 3: Write success measures before you launch
Direct answer: agree what will count as success at week 12 before anyone starts. Measure practice, capability and one business signal you already collect.
If you write the measures afterwards, you'll measure whatever looks good. Write them first, share them with your sponsor, and put them in the agreement with your vendor.
Success measures
| Layer | What to measure | Our target at week 12 | Where the data comes from |
|---|---|---|---|
| Practice | How many cohort managers practised at least one real conversation | Tool | |
| Practice | Which conversation types were practised, and when (before real meetings?) | Tool | |
| Capability | Share of practice conversations that reached the agreed standard | Tool | |
| Capability | How scores moved between first attempt and retry | Tool | |
| Capability | The most common gaps across the cohort | Tool (aggregate) | |
| Business signal | One measure you already collect: manager-related HR escalations, exit-interview themes, or engagement items about managers | HRIS / survey | |
| Manager voice | "I felt more prepared for the real conversation" (short pulse) | Pulse survey |
Template. Select the table and copy it. No email. Set your own targets. This guide does not invent benchmark numbers.
Pick one business signal, not five. Twelve weeks is short. A single signal you already track is more credible than a dashboard of new ones.
Phase 4: Launch around real moments (weeks 1–2)
Direct answer: launch when managers already have conversations coming up, with a short message from the sponsor and a clear privacy promise.
- Time it. A launch two to three weeks before a performance cycle or remuneration review gives managers an immediate reason to practise.
- Sponsor message first. A few lines from a senior leader about their own practice will do more than a polished campaign.
- One instruction, not a curriculum. "Before your next performance conversation, practise it once." That's the whole ask.
- Make access easy. If logging in is hard, the "once, before the meeting" moment is lost.
- Nominate a contact. One person in HR managers can ask, who isn't the person reviewing their performance.
Before you launch
- Problem sentence written
- Sponsor has practised and agreed to say so
- Privacy promise written and approved by your privacy officer
- Two or three conversations chosen
- First cohort named
- Success measures written and shared with the sponsor
- Launch timed to a real moment
- HR contact nominated
Phase 5: Support and review (weeks 3–12)
Direct answer: review at weeks 4, 8 and 12 against your written measures, and follow up with people, not dashboards.
Review rhythm
| When | Look at | Do |
|---|---|---|
| Week 4 | Who has practised; which conversations | Nudge managers with a known conversation coming up. Ask the sponsor to share a second story. |
| Week 8 | Capability movement; common gaps | Share one aggregate insight with the cohort (for example, "most of us forget to agree a date"). Consider adding a conversation. |
| Week 12 | All measures against your targets | Decide: expand, adjust or stop. Write down why. |
Common objections and honest answers
| Objection | Honest answer |
|---|---|
| "I don't have time." | You don't need a habit. Practise once, before the next conversation that matters. |
| "It'll feel fake." | The first minute might. A good counterpart pushes back, and that's usually when it stops feeling fake. |
| "HR will see what I say." | Point to the written privacy promise. If you can't make that promise, fix it before launch. |
| "AI doesn't understand our context." | The manager brings the context: the person, the situation, what happened. The tool provides the counterpart and the standard. |
| "I'm already good at this." | Then it'll be quick. Many experienced managers find a habit they didn't know they had, like burying the message. |
Phase 6: Decide whether to expand, adjust or stop
Direct answer: at week 12, compare results to the measures you wrote at the start, and make a written decision.
- Expand if managers practised before real conversations, capability moved and your business signal didn't go backwards. Add the next cohort and the next conversation.
- Adjust if practice was low but the people who used it improved. That's usually a timing or access problem, not a tool problem.
- Stop if managers practised and nothing moved, or if trust broke down. Write down what you learnt.
The decision itself is the deliverable. A 12-week cohort with a clear "here's what we saw and what we'll do next" is worth more than a year-long rollout nobody can evaluate.
Where Evro fits
Direct answer: Evro is built for this kind of rollout: a defined group of people leaders, the conversations they actually have, practice scored to a standard and success measures agreed in writing before you start.
For each conversation, managers in Evro:
- Choose the conversation they're about to have.
- Prepare a short plan using evidence-based frames (for performance feedback, SBI and Radical Candor).
- Roleplay with an AI counterpart who pushes back the way real people do.
- Get scored against a weighted rubric for that conversation, not on tone or filler words.
- Retry until they're at the line.
Practice isn't an HR file. Evro doesn't send session content or transcripts to HR. The organisation sees that critical conversations are being practised to a standard, not what anyone said.
Book a demo
Walk through a first-cohort plan and see a performance feedback conversation scored to a standard.
Book a demoFrequently asked questions
How long does it take to implement AI roleplay for managers?
A focused first cohort can be planned in a week or two and run for about 12 weeks. That is long enough for managers to practise real conversations across a performance or remuneration cycle and for HR to review results against measures agreed at the start.
How many managers should be in an AI roleplay pilot?
A first cohort of roughly 20 to 60 people leaders is usually enough to see patterns without a large change program. Include new and experienced managers and at least one senior leader who practises visibly first.
What should HR be able to see in an AI roleplay tool?
Decide before launch and put it in writing. A common, trust-preserving approach is that HR sees whether critical conversations are being practised and reaching the standard, but not what any manager said. If managers believe HR is reading their practice, they will not practise honestly.
How do you measure whether AI roleplay is working?
Track three layers: practice activity (who practised which conversations), capability (how many conversations reached the agreed standard and how scores moved on retry), and one business signal you already collect, such as HR escalations, exit-interview themes or engagement questions about managers.
Which conversations should managers practise first?
Start with two or three conversations managers will have this quarter. Performance feedback is usually the first. Pair it with coaching a high performer so the program is not seen as a performance-management tool, and add pay or promotion conversations ahead of remuneration review.
Do managers have time for AI roleplay?
Most managers will not practise daily. They will practise once, shortly before a conversation that matters. Design the rollout around that moment rather than around weekly practice targets.
What is Evro?
Evro is an AI platform for people leaders in Australian organisations. Managers prepare for a real conversation, such as performance feedback, rehearse it with an AI counterpart, are scored against a weighted rubric for that conversation and retry until they reach the standard. Practice sessions are not an HR file.
A last thought
Most of what makes a rollout work isn't in the software. It's a senior leader who goes first, a promise managers can trust and an invitation that arrives the week they actually need it.
Get those right, and a tool has a chance to become the place managers go before the conversations that matter. Get them wrong, and it might become one more login.
Book a demo
Walk through a first-cohort plan and see a performance feedback conversation scored to a standard.
Book a demoAbout Evro. Evro helps people leaders in Australian organisations rehearse the conversations that matter (performance feedback, coaching high performers, pay and promotion conversations, team conflict) and practise them to a standard before they happen. Managers prepare, roleplay with an AI counterpart, are scored against a rubric for that conversation and retry until they're ready. Practice sessions are not an HR file.